A r t i c l e s

A cascade of arabinosyltransferases controls shoot meristem size in tomato

Cao Xu1,9, Katie L Liberatore1,2,8,9, Cora A MacAlister1,8, Zejun Huang3,8, Yi-Hsuan Chu3, Ke Jiang1,8, Christopher Brooks1, Mari Ogawa-Ohnishi4, Guangyan Xiong5,8, Markus Pauly5,6, Joyce Van Eck7, Yoshikatsu Matsubayashi4, Esther van der Knaap3 & Zachary B Lippman1,2

Shoot meristems of plants are composed of stem cells that are continuously replenished through a classical feedback circuit involving the homeobox WUSCHEL (WUS) gene and the CLAVATA (CLV) gene signaling pathway. In CLV signaling, the CLV1 receptor complex is bound by CLV3, a secreted peptide modified with sugars. However, the pathway responsible for modifying CLV3 and its relevance for CLV signaling are unknown. Here we show that tomato inflorescence branching mutants with extra flower and fruit organs due to enlarged meristems are defective in arabinosyltransferase genes. The most extreme mutant is disrupted in a hydroxyproline O-arabinosyltransferase and can be rescued with arabinosylated CLV3. Weaker mutants are defective in arabinosyltransferases that extend arabinose chains, indicating that CLV3 must be fully arabinosylated to maintain meristem size. Finally, we show that a mutation in CLV3 increased fruit size during domestication. Our findings uncover a new layer of complexity in the control of plant stem cell proliferation.

A wide range of inflorescence architectures exist among closely phenotypes in maize and rice have shown that the CLV pathway related plants1, and this variation can be explained in part by the is conserved in dicots and monocots12. However, little is known rate at which meristems terminate in flowers and initiate new branch about the genes and molecular mechanisms underlying stem cell meristems from which new flowers will form2,3. However, less appre- proliferation in other crops, leaving unanswered whether the CLV ciated is the role of meristem size in shoot architecture, as well as pathway or other pathways are important in controlling meristem in flower and fruit production. In maize, for example, variation in size and productivity. Here we investigate the control of meristem

Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature meristem size can account for differences in kernel row number size in tomato. between inbred lines4, and high row number is a major driver of yield in cultivated hybrids5. Meristem size is controlled by a main- RESULTS tenance mechanism that replenishes stem cells lost to lateral organ Tomato fab and fin mutants have enlarged meristems npg formation6. In Arabidopsis thaliana, meristem maintenance is based We explored the genetic control of inflorescence architecture in on a feedback system comprising the stem cell–promoting WUS tomato by screening a large ethyl methanesulfonate (EMS) and fast homeodomain transcription factor and CLV signaling proteins. CLV neutron (FN) mutagenesis population for mutants with increased signaling involves binding of the glycopeptide CLV3 to cell surface inflorescence branching13. In contrast to previously characterized leucine-rich repeat (LRR) receptor complexes that include the recep- mutants in which unbranched wild-type inflorescences are trans- tor kinase CLV1 and related LRR receptors7–10. These interactions formed into highly branched structures with normal flowers as a result initiate a signaling cascade that restricts WUS expression to prevent of delay in meristem maturation3,14, we identified six new mutants stem cell overproliferation, and, through negative feedback, WUS where branching was associated with fasciated flowers having more promotes CLV3 expression to limit its own activity11. Mutations floral organs than wild-type flowers (Fig. 1a–h). Complementation in CLV pathway genes cause meristems to enlarge, which can lead tests identified two loci that we designated fasciated and branched to increased shoot and inflorescence branching, more flowers, and (fab: one EMS allele) and fasciated inflorescence (fin: two EMS and extra organs in flowers and fruits. Studies on mutants with these three FN alleles).

1Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, USA. 2Watson School of Biological Sciences, Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, USA. 3Department of Horticulture and Crop Science, Ohio State University, Wooster, Ohio, USA. 4Division of Biological Science, Graduate School of Science, Nagoya University, Chikusa, Nagoya, Japan. 5Energy Biosciences Institute, University of California, Berkeley, Berkeley, California, USA. 6Department of Plant and Microbial Biology, University of California, Berkeley, Berkeley, California, USA. 7Boyce Thompson Institute for Plant Science, Ithaca, New York, USA. 8Present addresses: Cereal Disease Laboratory, US Department of Agriculture, Agriculture Research Service, St. Paul, Minnesota, USA (K.L.L.), Department of Molecular, Cellular and Developmental Biology, University of Michigan, Ann Arbor, Michigan, USA (C.A.M.), Chinese Academy of Agricultural Sciences, Institute of Vegetables and Flowers, Beijing, China (Z.H.), Dow AgroSciences, LLC, Indianapolis, Indiana, USA (K.J.) and Department of Anatomical Sciences and Neurobiology, University of Louisville, Louisville, Kentucky, USA (G.X.). 9These authors contributed equally to this work. Correspondence should be addressed to Z.B.L. ([email protected]). Received 22 February; accepted 27 April; published online 25 May 2015; doi:10.1038/ng.3309

784 VOLUME 47 | NUMBER 7 | JULY 2015 Nature Genetics A r t i c l e s

Figure 1 The fab and fin mutants develop a b c d branched inflorescences with fasciated flowers WT fab fin fin as a consequence of enlarged meristems. (a) An unbranched inflorescence typical of a wild-type (WT) plant and a typical wild-type flower (inset). (b,c) Primary inflorescences from fab (b) and fin (c) mutants showing branching 1 1 1 8 2 11 and fasciated flowers (insets). (d) A branched 5 2 3 7 3 inflorescence from a fin side shoot. Red 4 9 5 4 3 6 5 7 Side-shoot in orescence arrowheads, branches; blue arrowheads, extra side shoots. (e–g) Wild-type fruits produce two e f g h locules (e), whereas the fruits of fab (f) and fin (g) WT fab fin WT fab fin mutants develop extra locules. Arrowheads, *** 16 *** locules. (h) Quantification and comparison of 14 *** *** 12 *** floral organ numbers in wild-type, fab and fin 10 *** *** flowers. (i–k) Stereoscope images of the primary 8 *** 6 SAM from wild-type (i), fab (j) and fin (k) plants 4 2

at the transition meristem (TM) stage, before Floral organ number 0 formation of the first flower. Dashed lines mark Sepal Petal Stamen Carpel the width and height used to measure meristem size. L8, leaf 8. (l) Quantification of TM size i j k l from wild-type, fab and fin plants. Data are WT fab fin 400 *** L8 means ± s.d.; n = 25 (h) and 9–13 (l). A two- 300 *** *** L8 L8 tailed, two-sample t test was performed, and 200 *** significant differences are represented by black asterisks: ***P < 0.0001. Scale bars: 1 cm (a–g) 100

and 100 µm (i–k). Meristem size ( µ m) 0 Width Height

Inflorescence branching in fin mutants was more severe than in and Supplementary Table 1). Notably, there was significant overlap fab mutants, and fin plants also developed a highly fasciated primary (P < 0.001) among the differentially expressed genes (Fig. 2a), sug- shoot with more side shoots than wild-type plants (Fig. 1a–d). Both gesting that the FAB and FIN genes could act in the same pathway. mutants produced larger fruits as a consequence of additional carpels, Because fab and fin meristems were enlarged, we checked for expres- with fin mutants bearing the largest fruits (Fig. 1e–g). We designated a sion changes in the homolog of WUS (SlWUS, Solyc02g083950)16,17 strong allele of fin as a reference (fin-e4489, hereafter fin unless other- and a likely functional homolog of CLV3 (SlCLV3, Solyc11g071380) wise noted), and quantification of floral organs in the corresponding (Supplementary Fig. 4a–c)18. Despite a 25-fold increase in SlCLV3 plants confirmed greater fasciation than in fab plants (Fig. 1h). expression, SlWUS expression doubled in fin meristems (Fig. 2b Analyses of double mutants of fab and fin with the meristem matura- and Supplementary Table 1). CLV-WUS feedback was also mildly tion mutants compound inflorescence (s) and anantha (an), defective affected in fab mutants: SlCLV3 expression was upregulated by Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature in the homolog of Arabidopsis WUSCHEL-RELATED HOMEOBOX 9 nearly fivefold, but SlWUS expression was maintained, consistent and the floral-identity gene UNUSUAL FLORAL ORGANS, respec- with a minor increase in vegetative meristem size (Supplementary tively, showed additive genetic relationships, indicating that the genes Fig. 2a–d and Supplementary Table 1). Notably, two additional npg underlying fin and fab act separately from the meristem maturation CLV3/EMBRYO-SURROUNDING REGION (CLE) genes (SlCLE3, pathway (Supplementary Fig. 1)14. Solyc02g067550; SlCLE9, Solyc06g074060) were also upregulated The phenotypes of fab and fin mutants were reminiscent of CLV in fin mutants (Fig. 2b and Supplementary Fig. 4d,e)18. We vali- gene mutants in Arabidopsis, in which inflorescence and flower fascia- dated the expression changes for SlWUS and SlCLV3 using in situ tion are caused by enlarged shoot apical meristems (SAMs)7,8,10,15. hybridization, which further showed substantial expansion of their The vegetative SAMs of the fab and fin mutants, unlike those of the s expression domains in fin meristems, consistent with stem cell and an mutants14, were larger than wild-type SAMs (Supplementary overproliferation (Fig. 2c). Fig. 2a–d), and this effect became more pronounced at the repro- ductive transition stage, just before formation of the first flower FAB encodes CLV1, and FIN encodes an arabinosyltransferase (Fig. 1i–l). We investigated meristem development in fin mutants We used a mapping strategy to clone the genomic regions underlying using scanning electron microscopy (SEM) and found that the the fab and fin mutants, using F2 populations derived from crosses enlarged SAM allowed more leaf primordia and thus axillary branch with the wild tomato Solanum pimpinellifolium (Fig. 2d,e and Online meristems to develop (Supplementary Fig. 2e–l). Methods). fab mapped to a 325-kb interval on the south arm of The phenotypic similarities between the fab and fin mutants and chromosome 4 where the closest homolog of CLV1 (Solyc04g081590) the fact that fab and fin double-mutant plants were more fasciated is located (Fig. 2d and Supplementary Fig. 5a). Like CLV1 than either single mutant alone (Supplementary Fig. 3) suggested (ref. 7), Solyc04g081590 was expressed in all tissue types, including a functional link in controlling meristem size. To better under- meristems, and sequencing its coding region in the fab mutant stand the relationship between these mutants, we compared RNA identified a missense mutation affecting the kinase domain (Fig. 2d,f sequencing profiles from fab, fin and wild-type vegetative primary and Supplementary Fig. 5b). Notably, the classical, weakly dominant- shoot apices (Online Methods). We detected a 2-fold or greater dif- negative clv1-9 allele encodes the same amino acid change19, and ference in expression for 679 genes in fab mutants and 1,194 genes plants heterozygous for the fab mutation exhibited weak fasciation in fin mutants in comparison to wild-type apices (Online Methods (Supplementary Fig. 6).

Nature Genetics VOLUME 47 | NUMBER 7 | JULY 2015 785 A r t i c l e s Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature Figure 2 Molecular characterization and positional cloning of the fin and fab mutants. (a) Venn diagram showing significant overlap of differentially expressed genes (P < 0.001) in wild-type versus fab (light gray) and wild-type versus fin (dark gray) vegetative apices by RNA sequencing. (b) Fold change in expression of SlCLV3, SlCLE3, SlCLE9 and SlWUS in fin and fab vegetative apices relative to wild-type apices by RNA sequencing (RNA-seq). npg (c) mRNA in situ hybridization showing substantial expansion in fin (right) and mild expansion in fab (center) of the SlWUS and SlCLV3 expression domains in vegetative meristems in comparison to wild type (left). (d,e) The mapping intervals for FAB (d) and FIN (e). The FAB and FIN genes are boxed in red. Protein diagrams are shown with characteristic domains and motifs labeled. Open arrows indicate point mutations and corresponding amino acid changes from fab-e0497 and two alleles of fin. Red horizontal lines in e represent three fin deletion alleles, which do not produce transcripts (Supplementary Fig. 5d). Rec., recombinant. (f,g) RNA sequencing read counts (RPKM) for FAB (f) and FIN (g) in major tissues, including the early vegetative meristem (EVM) and floral meristem (FM). (h) Subcellular colocalization of the transiently expressed FIN-GFP fusion protein with a cis-Golgi marker (Man49) in tomato protoplasts (Online Methods). Scale bars: 100 µm (c) and 10 µm (h).

fin mapped to a 1.1-Mb region in the middle of chromosome 11 found in the form of linear oligoarabinoside chains on extracellu- containing 71 genes. RNA sequencing identified a nonsense mutation lar Hyp-rich glycoproteins (HRGPs), such as cell wall–associated in Solyc11g064850, encoding a homolog of the recently identified extensins21–23. There are three HPAT genes in Arabidopsis, and hydroxyproline O-arabinosyltransferase (HPAT) protein family in extensins are underarabinosylated in hpat3 mutants and in hpat1 Arabidopsis (Fig. 2e and Supplementary Fig. 5c)20. Sanger sequencing and hpat2 double mutants20. Although cell wall thickness is reduced of Solyc11g064850 from four additional fin alleles found a mis- in hpat1 and hpat2 double-mutant plants, perhaps accounting for a sense mutation and deletions (Fig. 2e). RT-PCR confirmed that range of pleiotropic growth defects, none of the available HPAT gene Solyc11g064850 transcripts were absent in plants with the deletion mutants show defects in meristem size20. alleles (Supplementary Fig. 5d). Thus, FAB encodes the tomato FIN is one of four HPAT gene homologs in tomato and is most homolog of CLV1, and FIN encodes a predicted HPAT protein. similar to HPAT3 (Supplementary Fig. 5c). FIN is expressed in all HPATs are Golgi-localized type II transmembrane proteins that are tissue types (Fig. 2g), and transiently expressed FIN-GFP fusion pro- structurally similar to members of the GT8 glycosyltransferase fam- teins in tomato protoplasts colocalized with a Golgi marker (Fig. 2h). ily and catalyze the transfer of l-arabinose to the hydroxyl group of To determine whether altered extensin arabinosylation might explain hydroxyproline (Hyp) residues20. Hyp O-arabinosylation is primarily fin mutant phenotypes, we measured the monosaccharide composition

786 VOLUME 47 | NUMBER 7 | JULY 2015 Nature Genetics A r t i c l e s

Figure 3 Arabinosylated SlCLE peptides rescue fin meristem enlargement. sSICLV3(60 nM) SICLV3(60 nM) [Ara ]SICLV3(60 nM) a 3 (a) Representative images showing the SAMs of wild-type (top), fin (middle) and fab (bottom) plants treated with 60 nM of different versions of synthetic SlCLV3 peptide. Triarabinosylated SlCLV3 ([Ara3]SlCLV3) rescues the enlarged SAM of mutants, reducing it to a size comparable fin WT to that of wild-type SAMs. sSlCLV3, scrambled SlCLV3; SlCLV3, SlCLV3 peptide devoid of arabinosylation. Scale bars, 100 µm. (b) Quantification of wild-type, fin and fab SAM sizes after treatment with different versions of SlCLV3 and SlCLE9 peptides. The area of each SAM was captured and measured with ImageJ software (Online Methods). Data are presented as means ± s.d.; n = 30–46. All treatments were compared in a pairwise manner to the common control of wild type treated with scrambled fin SlCLV3 (Dunnett’s ‘compare with control’, ***P < 0.0001, **P < 0.001; NS, not significant).

of cell wall components isolated from the shoot apices of fin, fab fab and wild-type plants (Online Methods). We found no differences in arabinose content in extensin fractions (Supplementary Table 2). Moreover, transmission electron microscopy did not identify differ- sSICLV3 (60 nM) [Ara ]SICLV3 (60 nM) [Ara ]SICLE9 (60 nM) b 3 3 ences in cell wall thickness among fin, fab and wild-type meristems SICLV3 (60 nM) SICLE9 (60 nM) [Ara3]SICLV3 (30 nM) + 16,000 (Supplementary Fig. 7). Thus, meristem enlargement in fin mutants [Ara3]SICLE9 (30 nM) is not a result of changes in extensin arabinosylation or related cell *** *** wall defects. 14,000 ***

12,000 ** Engineered mutants of the tomato CLV pathway are fasciated 24

) *** Hyp O-arabinosylation is also found on secreted signaling peptides , 2 10,000 *** *** *** including the CLE peptide family9. CLE peptides are proteolytically *** *** processed from the C termini of pre-propeptides25. The mature 8,000 form of CLV3 is a 12- or 13-residue glycopeptide modified with

SAM area ( µ m 6,000 three l-arabinose residues on a Hyp residue at position 7 (Hyp7)9,26. NS NS NS NS Molecular modeling suggests that the triarabinoside chain induces NS 4,000 *** conformational changes that have important effects on binding and *** 26 specificity for receptor proteins . In support of this hypothesis, chem- 2,000 ically synthesized triarabinosylated CLV3 ([Ara3]CLV3) interacts in vitro with CLV1 more strongly than unmodified CLV3 (ref. 9) and 0 can rescue the enlarged meristems of clv3 mutants more effectively WT fin fab when applied exogenously26. Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature We tested whether SlCLV3 peptide controls meristem size in tomato this peptide is overexpressed in hpat3 mutants but not in hpat1 and by engineering null mutations in SlCLV3 using CRISPR/Cas9 gene- hpat2 double mutants, suggesting that HPATs are not equally respon- editing technology27–29. Using a CRISPR/Cas9 construct containing sible for arabinosylation of all CLE peptides20. As in vitro modifica- npg two single-guide RNAs (sgRNAs)29, we generated four first-generation tion assays using recombinantly expressed FIN or HPAT3 from yeast (T0) transgenic CRISPR (CR)-slclv3 plants that developed branched and tobacco BY-2 cell lines have been unable to detect arabinosyl- inflorescences with fasciated flowers, closely resembling fin mutants transferase activity on synthesized [Hyp7]SlCLV3 or [Hyp7]CLV3 (Online Methods and Supplementary Fig. 8a–d). PCR genotyping (data not shown and ref. 20), we treated live apices with CLE pep- and sequencing identified insertions and deletions (indels) in prox- tides (Online Methods). If loss of CLE arabinosylation causes fin imity to both sgRNA target sequences for all alleles (Online Methods mutant phenotypes, then arabinosylated SlCLE peptides should and Supplementary Fig. 8e). We also engineered mutations in FAB rescue the enlarged meristems. We synthesized [Ara3]SlCLV3 and and SlCLV2 (Solyc04g056640), a homolog of Arabidopsis CLV2 encod- [Ara3]SlCLE9 and grew seedlings in liquid culture in the presence ing an LRR receptor–like protein that acts redundantly with CLV1 of a low concentration of each peptide9,26. SAMs from fin seedlings 15,30 (Supplementary Fig. 5a) . CR-fab and CR-clv2 T0 plants were grown with 60 nM [Ara3]SlCLV3 or [Ara3]SlCLE9 were substantially weakly fasciated like fab mutants (Supplementary Figs. 9 and 10). reduced in size (Fig. 3a,b and Supplementary Table 4). This effect Notably, we validated fasciation for the CRISPR/Cas9 clv mutants in was greater with [Ara3]SlCLV3, resulting in a meristem size compa- T1 progeny plants (Supplementary Table 3). These findings demon- rable to that of wild-type plants treated with non-arabinosylated or strate that the function of the clv meristem maintenance pathway is scrambled SlCLV3. Notably, combining 30 nM [Ara3]SlCLV3 with conserved in tomato. 30 nM [Ara3]SlCLE9 resulted in a rescue of fin meristem size simi- lar to that observed with 60 nM [Ara3]SlCLV3, implying a role for Arabinosylated SlCLV3 can rescue fin mutants SlCLE9 in CLV signaling (Fig. 3b and Supplementary Table 4). fab The enlarged meristems of fin mutants are in striking contrast to meristems were insensitive to all [Ara3]SlCLE treatments, consistent the dramatic upregulation of SlCLE genes, suggesting that a defect with a defect in transducing CLE peptide signals25. These results in CLV signaling may be due to a failure to post-translationally support the model that stem cell overproliferation in fin meristems arabinosylate SlCLV3 and/or other CLE peptides. In support of this is due to a break in the CLV-WUS circuit, caused by a failure to form hypothesis, arabinosylation of Arabidopsis CLE2 is reduced when triarabinosylated CLE peptides.

Nature Genetics VOLUME 47 | NUMBER 7 | JULY 2015 787 A r t i c l e s

Figure 4 The fab2 mutant is defective in an a b WT fab2 arabinosyltransferase gene predicted to extend 14 fab2 *** *** ** *** arabinose chains. (a) Primary inflorescence 12 from a fab2 mutant and a representative flower 10 (inset). Red arrowheads mark branches. 8 (b) Quantification and comparison of floral 6 organ numbers in fab2 and wild-type plants. 4 2 (c) Stereoscope images of the primary SAM Floral organ number of fab2 plants, showing an enlarged TM stage 0 Sepal Petal Stamen Carpel in comparison to wild type. Dashed lines mark the height and width used for meristem c d 300 size quantification. (d) Quantification and * WT WT 250 fab2 comparison of TM meristem sizes in fab2 and wild-type plants. (e) Positional cloning of FAB2. 200 The FAB2 gene is boxed in red, and a protein 150 * 1 diagram is shown below. Open arrows mark the 8 2 fab2 100 3 nonsense and missense point mutations from 50 7 Meristerm size ( µ m) two independently derived alleles of fab2. 4 0 6 (f) RNA sequencing counts (RPKM) of FAB2 in 5 Width Height major tissue types. (g) Subcellular colocalization FAB2 2,500 of transiently expressed FAB2-GFP fusion e FAB2 (XEG113) f protein with a medial-Golgi marker (Mur3) in (Solyc08g075340) 2,000 Chr. 8 tomato protoplasts (Online Methods). (h) qRT- 10 kb 1,500 PCR showing upregulation of SlWUS, SlCLV3

RPKM 1,000 and SICLE9 in fab2 vegetative apices (Online 1 rec. FAB2 1 rec. Methods). Data are means s.d.; = 13 (b), 500 ± n N- TM Reticulon Arabinosyltransferase Reticulon 635 aa 8 (d) and 3 (h). A two-tailed, two-sample t test LRRs ^ ^ 0 Q118* L332P was performed, and significant differences are FM fab2-e0051 fab2-e4981 EVM Root Leaf represented by black asterisks: ***P < 0.0001, Flower **P < 0.001, *P < 0.01. Scale bars: 1 cm (a), g h SlWUS SlCLV3 SICLE9 Medial- 2.0 20 14 100 µm (c) and 10 µm (g). FAB2-GFP Golgi–mCherry Merged Bright field 12 1.6 16 10 1.2 12 8 0.8 8 6 Loss of an arabinosyltransferase cascade 4 0.4 4 causes fasciation 2 (normalized to WT) Relative transcripts 0 0 0 By rescreening the original mutagenesis WT WT WT population13 and a new EMS-mutagenized fab2 fab2 fab2 population, we identified two alleles for fab2, a mildly fasciated and branched mutant that closely resembled developed branched inflorescences with fasciated flowers, with these fab mutants (Fig. 4a–d). Positional cloning of fab2 identified a mis- phenotypes more severe than those of the fab2 mutants, and sequencing Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature sense and nonsense mutation in the tomato homolog of Arabidopsis identified a range of indel mutations for all alleles, including a XYLOGLUCANASE 113 (XEG113), encoding a predicted arabinosyl- homozygous small deletion in one plant (Fig. 5d–h). Using quanti- transferase in the GT77 family (Fig. 4e and Supplementary Fig. 11). tative RT-PCR (qRT-PCR), we found that SlWUS, SlCLV3 and SlCLE9 npg Arabidopsis plants with mutations in XEG113 display predominantly were all upregulated, with SlCLV3 and SlCLE9 showing more than unusual diarabinoside chains on extensins, resulting in longer 15-fold higher expression than in wild type (Fig. 5i). Thus, FIN, hypocotyls and shorter root hairs31,32. In tomato, FAB2 is expressed RRA3a and FAB2 all function in the tomato CLV pathway. in all tissues and localizes to Golgi like FIN (Fig. 4f,g). Similar to in fin mutants, the expression levels of SlWUS, SlCLV3 and SlCLE9 were all A natural CLV3 mutation was selected during domestication upregulated in fab2 SAMs, with SlCLV3 and SlCLE9 showing greater Our CLV pathway mutants reminded us of naturally occurring fasciation than tenfold upregulation in their expression (Fig. 4h). Thus, FAB2, responsible for the dramatic increase in tomato fruit size during like FIN, is a critical regulator of CLV signaling. domestication34. Extreme size is based on more carpels, which Mutations in Arabidopsis REDUCED RESIDUAL ARABINOSE 3 increases the number of seed compartments (locules). In com- (RRA3), also encoding an arabinosyltransferase in the GT77 fam- parison to the invariant bilocular fruits of wild tomatoes and most ily33, result in shorter root hairs as a consequence of predominantly small-fruited cultivars, large-fruited varieties can have eight or more monoarabinoside chains on extensins32. This observation suggested locules34. A large portion of this variation is due to locule number that RRA3 and XEG113 act sequentially after HPATs to extend ara- (lc) and fasciated (fas), two classical mutations that have synergis- binoside chains20,32. To test whether the tomato homolog of RRA3 tic effects on locule number and thus fruit size when combined35,36. functions with FIN and FAB2 in tomato CLV signaling, we mutated The effect in lc mutants is weaker than in fas mutants and is likely due the closest homolog of RRA3 (Solyc04g080080) using CRISPR/Cas9 to a regulatory mutation downstream of SlWUS16, which has been (Fig. 5 and Supplementary Fig. 12)29. We identified two RRA genes proposed to affect binding of the tomato homolog of AGAMOUS, in tomato; both were most similar to RRA3, but only Solyc04g080080 a transcription factor that downregulates WUS in Arabidopsis floral (hereafter referred to as SlRRA3a) encoded a predicted transmem- meristems37. The mutant fas phenotype was reported to be caused by brane domain (Fig. 5a). SlRRA3a is expressed in all tissues, and tran- loss of expression of a YABBY transcription factor gene38. However, siently expressed SlRRA3a-GFP fusion proteins in tomato protoplasts the YABBY gene (Solyc11g071810) is located on chromosome 11 close colocalized with a Golgi marker (Fig. 5b,c). Eight CR-slrra3a plants to SlCLV3, and further dissection of the fas locus identified a 294-kb

788 VOLUME 47 | NUMBER 7 | JULY 2015 Nature Genetics A r t i c l e s

a c f SlRRA3a SIRRA3a- Trans- (Solyc04g080080) GFP Golgi–mCherry Merged Bright field CR-slrra3a-4 N- Arabinosyltransferase 428 aa Transmembrane domain b SlRRA3a 700 d SlRRA3a 46 bp 600 ATG F Target 1 Target 2 Stop 500 R 400 e WT 1 2 3 4 5 6 7 8 9 10 11 12 WT 300 RPKM bp SlRRA3a 300 200 2 200 4 100 Cas9 0 10 6 Fasciation ++ + + + + + + FM 8 EVM Root Leaf Flower

g WT CR-slrra3a h Target 1 PAM Target 2 PAM *** *** SLRRA3a 12 *** 10 *** a1 CR-slrra3a-2 a2 8 a3 6 CR-slrra3a-4 4 2 a1 Floral organ number CR-slrra3a-5 a2 0 a3 Sepal Petal Stamen Carpel i SlWUS 30 SlCLV324 SlCLE9 Figure 5 CRISPR/Cas9-engineered mutations in the arabinosyltransferase gene SlRRA3a result 6 in fasciated plants. (a) SlRRA3a protein diagram. (b) RNA sequencing counts (RPKM) for SlRRA3a 4 20 16 in major tissues. (c) Subcellular colocalization of SlRRA3a-GFP with a trans-Golgi marker (ST) in tomato protoplasts (Online Methods). (d) Schematic illustrating two sgRNAs (red arrows) targeting 2 10 8 SlRRA3a. Black arrows represent PCR genotyping primers. (e) PCR genotyping of 12 first-generation (normalized to WT) Relative transcripts 0 0 0 (T0) CR-slrra3a plants showing indels. PCR of Cas9 is shown along with presence of fasciation.

(f) Inflorescence from a CR-slrra3a plant showing branching (red arrowheads) and a fasciated flower WT WT WT (inset). (g) Quantification and statistical comparison of floral organ numbers in wild-type and CR-slrra3a plants. Data were collected from plants with confirmed mutations in all sequenced alleles (h; see also CR-slrra3a CR-slrra3a CR-slrra3a Online Methods). (h) CR-slrra3a alleles identified from three T0 plants. CR-slrra3a-2 and CR-slrra3a-5 plants were chimeric, with multiple indels (blue dashed lines and letters). One of each representative allele identified by sequencing is shown. The CR-slrra3a-4 plant was homozygous for a 2-nt deletion. Red font highlights sgRNA targets, and black boxes indicate protospacer-adjacent motif (PAM) sequences; a, allele. (i) qRT-PCR showing upregulation of SlWUS, SlCLV3 and SICLE9 in CR-slrra3a plants. RNA was extracted from the side- shoot apices of wild-type and CR-slrra3a T plants (Online Methods). Data are means ± s.d.; n = 11 (g) and 3 (i). A two-tailed, two-sample t test was

Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature 0 performed, and significant differences are represented by black asterisks: ***P < 0.0001. Scale bars: 10 µm (c) and 1 cm (f).

npg inversion with breakpoints in intron 1 of YABBY and 1 kb upstream did not show reduced locule number (Fig. 6f). Notably, we validated 39 of SlCLV3 (Fig. 6a,b) . This finding, in combination with weak these effects in T1 long-gCLV3 progeny plants and found that rescue rescue of the fas mutant upon introduction of functional YABBY38, of the fas phenotype cosegregated with the transgene (Fig. 6f). led us to ask whether SlCLV3 might also contribute to naturally We therefore conclude that a weak regulatory mutation in SlCLV3 occurring fasciation. underlies the fas mutant. Introgression of the fas locus into our standard cultivated background resulted in weak branching and flower and fruit fas- DISCUSSION ciation resembling the phenotype for CR-clv2 plants (Fig. 6c and By uniting forward genetics with CRISPR/Cas9 technology27, we Supplementary Fig. 10). Introgression of fas into S. pimpinellifolium exposed arabinosyltransferase genes as critical new components of the (Sp-fas) produced even weaker fasciation, with little branching and CLV pathway that are essential to control meristem size. As members half of all fruits producing three locules, likely owing to the absence of the plant and animal glycosyltransferase superfamily40–42, several of modifier loci in the wild species background37. Such a weak effect putative arabinosyltransferases have recently been found to function would be consistent with a change in SlCLV3 expression38, and in diverse processes, including pollen tube growth by HPATs20, root in situ hybridization showed reduced expression of SlCLV3 in the floral hair elongation by RRAs and XEG113 (ref. 32), and Rhizobia-induced meristems of Sp-fas plants (Fig. 6d). To determine whether SlCLV3 nodulation by FIN homologs in pea and clover43. Yet, no meristem underlies the fas mutant, we carried out transgenic complementation phenotypes for any available arabinosyltransferase mutants have been experiments in the Sp-fas genotype (Online Methods). We generated reported. Our findings that arabinosyltransferase genes are critical 18 plants carrying a construct comprising the SlCLV3 gene with for stem cell homeostasis suggest that members of different arabino- 5.5 kb of upstream and 3.5 kb of downstream regions (long-gCLV3), syltransferase families were coopted to target distinct sets of proteins and all plants were rescued to bilocular, smaller fruits (Fig. 6e). In and signaling peptides in different species and developmental con- contrast, 16 plants transformed with a similar construct containing texts. In support of this hypothesis, unlike the predominant fasciation 1 kb upstream of SlCLV3 (up to the inversion breakpoint; short-gCLV3) phenotype of fin mutants, we found that fab2 and CR-rra3a plants

Nature Genetics VOLUME 47 | NUMBER 7 | JULY 2015 789 A r t i c l e s

a b WT fas c d Chr. 11: 54.88 Mb 294-kb inversion 55.17 Mb SlCLV3 P1 + P2 P3 + P4 P1 + P3 P2 + P4 P1 + P2 P3 + P4 P1 + P3 P2 + P4 fas

SlCLV3 YABBY 42 genes WT P1 P2 P3 P4 e WT SlCLV3 P1 P3 P2 P4 fas

6 5 1 Sp-fas 4 Figure 6 The fasciated (fas) locus, which led to WT Sp-fas + 2 long-gSlCLV3 3 Sp-fas increased fruit size during tomato domestication, resulted from a regulatory mutation in SlCLV3. f 4 (a) The locus is a 294-kb inversion with breakpoints in the first intron of a YABBY transcription factor fas ** ** gene and 1 kb upstream of the SlCLV3 start codon38,39. (b,c) Introgression of the fas locus into the 3 ** cultivar M82, confirmed by PCR (b), results in weak inflorescence branching (c, red arrowheads) and 39 flower fasciation. P1–P4, primers used to detect the left and right borders of the inversion 2 (Supplementary Table 7). (d) In situ hybridization showing reduced SlCLV3 expression in the floral meristems of plants with the locus introgressed ( ) as compared to S. pimpinellifolium fas Sp-fas Locule number wild-type plants. (e) S. pimpinellifolium plants invariantly produce bilocular fruits (left), whereas half of 1 Sp-fas fruits have three locules (middle); the mutant phenotype is rescued with a SlCLV3 genomic transgene (long-gCLV3; right) (Online Methods). (f) Quantification of locule number in first-generation (T0) Sp-fas 0 transgenic plants carrying long-gCLV3 in comparison to plants carrying a similar construct having only WT 1 kb upstream of SlCLV3 (up to the inversion breakpoint; short-gCLV3). Data to the right of the vertical dashed Sp-fas line are from T progeny plants that carry the long-gCLV3 transgene or lack the transgene, showing that rescue 1 long-gSlCLV3 short-gSlCLV3 long-gSlCLV3 0 1 non-transgenic of the fas mutant cosegregates with the presence of the transgene (Online Methods). Data are means ± s.d.; T T 0 T T 1 n ≥ 200. A two-tailed, two-sample t test was performed, and significant differences are represented by black asterisks: **P < 0.001. Scale bars: 1 cm (c,e) and 100 µm (d).

also suffer from sterility defects (data not shown). In the context of genes that can be targeted to achieve this control will extend to of stem cell maintenance, our data imply that addition of complete other LRR receptor–like genes and CLE genes operating in parallel and arabinose chains on CLV3 and related CLE peptides by a cascade compensatory pathways45–50 (Fig. 3). In this respect, the upregulation of arabinosyltransferases is required to fully activate the conserved of multiple potentially functionally redundant SlCLE genes when FIN CLV-WUS circuit. Notably, in this respect, fasciation in fab2 mutants is mutated might explain the extreme fasciation in fab and fin double is weaker than in fin mutants and CR-rra3a plants appear intermediate mutants, if higher levels of unmodified SlCLE peptides provide some (Supplementary Table 5), suggesting that the potency of CLE peptide activity through SlCLV1 (Fig. 2b and Supplementary Fig. 3). Notably, signaling in planta increases with arabinose chain length26. Thus, our results suggest that the ability to fine-tune CLV signaling could Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature developmentally programmed or environmentally induced changes potentially be exploited to improve crop productivity. For example, in in arabinosyltransferase activity on CLV3 and related CLE peptides parallel to the improvement of tomato fruit size by fas, induced weak could be a mechanism to quantitatively control stem cell proliferation alleles of a CLV2 homolog in maize can increase kernel row number4, npg and meristem size. and a naturally occurring CLV3 mutation in mustard (Brassica rapa) Our findings show that disrupting the tomato CLV-WUS circuit could provide increased seed production51. With the widespread suc- has dramatic consequences for inflorescence architecture, flower pro- cess of the CRISPR/Cas system in plants, the entire CLV pathway, now duction and fruit size. Increased fruit size was likely the first stage of including arabinosyltransferase genes, can be targeted in many crops tomato domestication, and, although many loci were involved34, the to test whether customized alleles can benefit breeding. synergistic effect of fas and lc mutations on flower and fruit fascia- tion may have largely been responsible for fruits that exceed 500g URLs. Solanaceae Genomics Network, http://solgenomics.net/; (refs. 35,36). In support of this idea, combining fas with lc in Solanaceae Genomics Network FTP site for genomic data deposi- S. pimpinellifolium enhances fasciation (Supplementary Fig. 13). tion and access, ftp://ftp.solgenomics.net/transcript_sequences/by_ Our findings that fas is due to partial loss of SlCLV3 expression and experiment/lippman_lab/Liberatore_etal/; tomato tissue RNA sequencing that a regulatory change in SlWUS is the likely cause of lc16 indicate database, http://tomatolab.cshl.edu/~lippmanlab2/allexp_query.html; that tomato domestication relied on subtle changes in the activity of ImageJ software, http://rsb.info.nih.gov/ij/download.html. the CLV-WUS circuit. Although the CLV-WUS circuit in Arabidopsis seems to be buffered against large changes in CLV3 expression44, Methods fas mutants demonstrate that even a modest decrease in expression Methods and any associated references are available in the online can cause a substantial increase in fruit size, without compromising version of the paper. plant growth and architecture. In this respect, it is notable that our collection of CLV pathway mutations provided a range of fasciation, Accession codes. RNA sequencing data have been deposited to indicating that there are multiple ways to manipulate the CLV-WUS the Solanaceae Genomics Network (SGN; http://solgenomics.net/). circuit to quantitatively control meristem size and, thus, inflorescence Data are available from the following ftp link: ftp://ftp.solgenomics.net/ architecture, flower production and fruit size. We predict that the set transcript_sequences/by_experiment/lippman_lab/Liberatore_etal/.

790 VOLUME 47 | NUMBER 7 | JULY 2015 Nature Genetics A r t i c l e s

Note: Any Supplementary Information and Source Data files are available in the 17. Reinhardt, D., Frenz, M., Mandel, T. & Kuhlemeier, C. Microsurgical and laser online version of the paper. ablation analysis of interactions between the zones and layers of the tomato shoot apical meristem. Development 130, 4073–4083 (2003). 18. Zhang, Y., Yang, S., Song, Y. & Wang, J. Genome-wide characterization, expression Acknowledgments and functional analysis of CLV3/ESR gene family in tomato. BMC Genomics 15, We thank members of the Lippman laboratory, especially S. Thomain for her initial 827 (2014). finding of FAB2 and for invaluable conversations that helped shape this work. 19. Diévart, A. et al. CLAVATA1 dominant-negative alleles reveal functional overlap We thank D. Zamir (Hebrew University of Jerusalem) for providing mutants between multiple receptor kinases that regulate meristem and organ development. and also Y. Eshed (Weizmann Institute of Science) for providing mutants and Plant Cell 15, 1198–1211 (2003). comments on the manuscript. We thank S. Hearn at the Cold Spring Harbor 20. Ogawa-Ohnishi, M., Matsushita, W. & Matsubayashi, Y. Identification of three Laboratory St. Giles Advanced Microscopy Center for providing technical service hydroxyproline O-arabinosyltransferases in Arabidopsis thaliana. Nat. Chem. Biol. for transmission electron microscopy. We thank W. Wang for assistance with tomato 9, 726–730 (2013). 21. Kieliszewski, M.J. & Lamport, D.T. Extensin: repetitive motifs, functional sites, transformation, X. Song for advice on peptide assays, DuPont Pioneer for research post-translational codes, and phylogeny. Plant J. 5, 157–172 (1994). support, and T. Mulligan, A. Krainer and staff from Cornell University’s Long 22. Lamport, D.T., Kieliszewski, M.J., Chen, Y. & Cannon, M.C. Role of the Island Horticultural Research and Extension Center in Riverhead, New York, for extensin superfamily in primary cell wall architecture. Plant Physiol. 156, 11–19 assistance with plant care. This research was supported by the Energy Biosciences (2011). Institute and the Fred Dickinson Chair for M.P., a National Science Foundation 23. Cannon, M.C. et al. Self-assembly of the plant cell wall requires an extensin scaffold. Graduate Research Fellowship (DGE-0914548) to K.L.L., a Gordon and Betty Moore Proc. Natl. Acad. Sci. USA 105, 2226–2231 (2008). Foundation Fellowship from the Life Sciences Research Foundation to C.A.M., 24. Matsubayashi, Y. Small post-translationally modified peptide signals in.Arabidopsis. grants from the National Science Foundation Plant Genome Research Program to Arabidopsis Book 9, e0150 (2011). 25. Xu, T.T., Song, X.F., Ren, S.C. & Liu, C.M. The sequence flanking the N-terminus E.v.d.K. (0922661) and to J.V.E. and Z.B.L. (1237880), and an Agriculture and Food of the CLV3 peptide is critical for its cleavage and activity in stem cell regulation Research Initiative competitive grant (2015-67013-22823) of the US Department of in Arabidopsis. BMC Plant Biol. 13, 225 (2013). Agriculture National Institute of Food and Agriculture to Z.B.L. 26. Shinohara, H. & Matsubayashi, Y. Chemical synthesis of Arabidopsis CLV3 glycopeptide reveals the impact of hydroxyproline arabinosylation on peptide AUTHOR CONTRIBUTIONS conformation and activity. Plant Cell Physiol. 54, 369–374 (2013). C.X., K.L.L., C.A.M., Z.H., Y.-H.C., M.P., J.V.E., Y.M., E.v.d.K. and Z.B.L. designed 27. Doudna, J.A. & Charpentier, E. Genome editing. The new frontier of genome engineering with CRISPR-Cas9. Science 346, 1258096 (2014). and planned experiments. C.X., K.L.L., C.A.M., Z.H., Y.-H.C., C.B., M.O.-O., G.X. 28. Belhaj, K., Chaparro-Garcia, A., Kamoun, S., Patron, N.J. & Nekrasov, V. Editing and Z.B.L. performed experiments and collected the data. C.X., K.L.L., C.A.M., plant genomes with CRISPR/Cas9. Curr. Opin. Biotechnol. 32, 76–84 (2015). Z.H., Y.-H.C., K.J., C.B., M.O.-O., G.X., M.P., Y.M., E.v.d.K. and Z.B.L. analyzed the 29. Brooks, C., Nekrasov, V., Lippman, Z.B. & Van Eck, J. Efficient gene editing in data. K.L.L., C.X., Y.M., E.v.d.K. and Z.B.L. designed the research. C.X., K.L.L. and tomato in the first generation using the clustered regularly interspaced short Z.B.L. wrote the manuscript. palindromic repeats/CRISPR-associated9 system. Plant Physiol. 166, 1292–1297 (2014). COMPETING FINANCIAL INTERESTS 30. Nimchuk, Z.L., Tarr, P.T. & Meyerowitz, E.M. An evolutionarily conserved The authors declare competing financial interests: details are available in the online pseudokinase mediates stem cell production in plants. Plant Cell 23, 851–854 (2011). version of the paper. 31. Gille, S., Hansel, U., Ziemann, M. & Pauly, M. Identification of plant cell wall mutants by means of a forward chemical genetic approach using hydrolases. Reprints and permissions information is available online at http://www.nature.com/ Proc. Natl. Acad. Sci. USA 106, 14699–14704 (2009). reprints/index.html. 32. Velasquez, S.M. et al. O-glycosylated cell wall proteins are essential in root hair growth. Science 332, 1401–1403 (2011). 1. Rickett, H.W. The classification of inflorescences. Bot. Rev. 10, 187–231 33. Egelund, J. et al. Molecular characterization of two Arabidopsis thaliana (1944). glycosyltransferase mutants, rra1 and rra2, which have a reduced residual arabinose 2. Prusinkiewicz, P., Erasmus, Y., Lane, B., Harder, L.D. & Coen, E. Evolution and content in a polymer tightly associated with the cellulosic wall residue. Plant Mol. development of inflorescence architectures. Science 316, 1452–1456 (2007). Biol. 64, 439–451 (2007). 3. Park, S.J., Eshed, Y. & Lippman, Z.B. Meristem maturation and inflorescence 34. Tanksley, S.D. The genetic, developmental, and molecular bases of fruit size and architecture—lessons from the Solanaceae. Curr. Opin. Plant Biol. 17, 70–77 shape variation in tomato. Plant Cell 16 (suppl.), S181–S189 (2004).

Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature (2014). 35. Barrero, L.S. & Tanksley, S.D. Evaluating the genetic basis of multiple-locule fruit 4. Bommert, P., Nagasawa, N.S. & Jackson, D. Quantitative variation in maize kernel in a broad cross section of tomato cultivars. Theor. Appl. Genet. 109, 669–679 row number is controlled by the FASCIATED EAR2 locus. Nat. Genet. 45, 334–337 (2004). (2013). 36. Lippman, Z. & Tanksley, S.D. Dissecting the genetic pathway to extreme fruit size 5. Doebley, J. The genetics of maize evolution. Annu. Rev. Genet. 38, 37–59 (2004). in tomato using a cross between the small-fruited wild species Lycopersicon npg 6. Barton, M.K. Twenty years on: the inner workings of the shoot apical meristem, a pimpinellifolium and L. esculentum var. Giant Heirloom. Genetics 158, 413–422 developmental dynamo. Dev. Biol. 341, 95–113 (2010). (2001). 7. Clark, S.E., Williams, R.W. & Meyerowitz, E.M. The CLAVATA1 gene encodes a 37. van der Knaap, E. et al. What lies beyond the eye: the molecular mechanisms putative receptor kinase that controls shoot and floral meristem size in Arabidopsis. regulating tomato fruit weight and shape. Front. Plant Sci. 5, 227 (2014). Cell 89, 575–585 (1997). 38. Cong, B., Barrero, L.S. & Tanksley, S.D. Regulatory change in YABBY-like 8. Fletcher, J.C., Brand, U., Running, M.P., Simon, R. & Meyerowitz, E.M. Signaling transcription factor led to evolution of extreme fruit size during tomato domestication. of cell fate decisions by CLAVATA3 in Arabidopsis shoot meristems. Science 283, Nat. Genet. 40, 800–804 (2008). 1911–1914 (1999). 39. Huang, Z. & van der Knaap, E. Tomato fruit weight 11.3 maps close to 9. Ohyama, K., Shinohara, H., Ogawa-Ohnishi, M. & Matsubayashi, Y. A glycopeptide fasciated on the bottom of chromosome 11. Theor. Appl. Genet. 123, 465–474 regulating stem cell fate in Arabidopsis thaliana. Nat. Chem. Biol. 5, 578–580 (2009). (2011). 10. Stahl, Y. & Simon, R. Plant primary meristems: shared functions and regulatory 40. Lombard, V., Golaconda Ramulu, H., Drula, E., Coutinho, P.M. & Henrissat, B. The mechanisms. Curr. Opin. Plant Biol. 13, 53–58 (2010). carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res. 42, 11. Schoof, H. et al. The stem cell population of Arabidopsis shoot meristems in D490–D495 (2014). maintained by a regulatory loop between the CLAVATA and WUSCHEL genes. 41. Yin, Y., Mohhen, D., Gelineo-Albersheim, I., Xu, Y. & Hahn, M. in Annual Plant Cell 100, 635–644 (2000). Reviews: Plant Polysaccharides: Biosynthesis and Bioengineering (ed. Ulvskov, P.) 12. Pautler, M., Tanaka, W., Hirano, H.Y. & Jackson, D. Grass meristems I: shoot apical (Wiley-Black, 2011). meristem maintenance, axillary meristem determinacy and the floral transition.Plant 42. Petersen, B.L., Faber, K. & Ulvskov, P. in Annual Plant Reviews: Plant Cell Physiol. 54, 302–312 (2013). Polysaccharides: Biosynthesis and Bioengineering (ed. Ulvskov, P.) (Wiley-Black, 13. Menda, N., Semel, Y., Peled, D., Eshed, Y. & Zamir, D. In silico screening of a 2011). saturated mutation library of tomato. Plant J. 38, 861–872 (2004). 43. Schnabel, E.L. et al. The ROOT DETERMINED NODULATION1 gene regulates 14. Lippman, Z.B. et al. The making of a compound inflorescence in tomato and related nodule number in roots of Medicago truncatula and defines a highly conserved, nightshades. PLoS Biol. 6, e288 (2008). uncharacterized plant gene family. Plant Physiol. 157, 328–340 (2011). 15. Jeong, S., Trotochaud, A.E. & Clark, S.E. The Arabidopsis CLAVATA2 gene encodes 44. Müller, R., Borghi, L., Kwiatkowska, D., Laufs, P. & Simon, R. Dynamic and a receptor-like protein required for the stability of the CLAVATA1 receptor-like kinase. compensatory responses of Arabidopsis shoot and floral meristems to CLV3 signaling. Plant Cell 11, 1925–1934 (1999). Plant Cell 18, 1188–1198 (2006). 16. Muños, S. et al. Increase in tomato locule number is controlled by two single- 45. Betsuyaku, S. et al. Mitogen-activated protein kinase regulated by the CLAVATA nucleotide polymorphisms located near WUSCHEL. Plant Physiol. 156, 2244–2254 receptors contributes to shoot apical meristem homeostasis. Plant Cell Physiol. 52, (2011). 14–29 (2011).

Nature Genetics VOLUME 47 | NUMBER 7 | JULY 2015 791 A r t i c l e s

46. DeYoung, B.J. et al. The CLAVATA1-related BAM1, BAM2 and BAM3 receptor 49. Müller, R., Bleckmann, A. & Simon, R. The receptor kinase CORYNE of Arabidopsis kinase-like proteins are required for meristem function in. Arabidopsis. Plant J. 45, transmits the stem cell–limiting signal CLAVATA3 independently of CLAVATA1. Plant 1–16 (2006). Cell 20, 934–946 (2008). 47. DeYoung, B.J. & Clark, S.E. BAM receptors regulate stem cell specification and 50. Nimchuk, Z.L., Zhou, Y., Tarr, P.T., Petersen, B.A. & Meyerowitz, E.M. Plant stem organ development through complex interactions with CLAVATA signaling. Genetics cell maintenance by transcriptional cross-regulation of related receptor kinases. 180, 895–904 (2008). Development 142, 1043–1049 (2015). 48. Kinoshita, A. et al. RPK2 is an essential receptor-like kinase that transmits the 51. Fan, C. et al. A novel single-nucleotide mutation in a CLAVATA3 gene homolog controls CLV3 signal in Arabidopsis. Development 137, 3911–3920 (2010). a multilocular silique trait in Brassica rapa L. Mol. Plant 7, 1788–1792 (2014). Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature npg

792 VOLUME 47 | NUMBER 7 | JULY 2015 Nature Genetics ONLINE METHODS potassium ferrocyanide. Samples were dehydrated in a graded alcohol series Plant materials and growth conditions. Six branched and fasciated inflo- followed by 100% dry acetone. Samples were infiltrated with agitation in 50% rescence mutants forming two complementation groups were isolated from a Epon Araldite resin in 100% acetone for 2 h and then overnight in 100% saturated mutagenesis population in the standard genetic background cultivar resin. Samples were flat embedded in resin in BEEM capsules from which M82 (ref. 13). One complementation group, comprising a single EMS mutant, the conical ends had been removed and were capped with the flat lid of the e0497, was designated fasciated and branched (fab). The other five mutants capsule. Plastic blocks were carefully trimmed and positioned for sectioning (three EMS and two FN alleles: fin-e4489, fin-e4643, fin-e9501, fin-n2326 to obtain longitudinal sections through the meristem region. Thin sections and fin-n5644) were designated fasciated inflorescence (fin). We rescreened were collected on Butvar-coated 1 × 2 mm nickel slot grids (Veco, Electron the mutagenesis population and, using the same EMS mutagenesis protocol, Microscopy Sciences) and counterstained in lead citrate stain. Samples were developed and screened a new population of 3,000 M2 families. We identified examined with a Hitachi H700 transmission electron microscope at 75 kV and two alleles of a new mutant phenotypically similar but not complementary were recorded on Kodak 4489 film that was then scanned at 2,400 dpi using to fab that was designated fab2 (fab2-e4891 and fab2-e0051). Three reference an Epson Perfection V750 Pro scanner. alleles, fab-e0497, fin-e4489 and fab2-e4891, were backcrossed to the M82 parental line at least three times before generating F2 mapping populations Meristem transcriptome profiling. Vegetative meristem collection, RNA extraction or double mutants. F2 mapping populations were generated by crossing ref- and library preparation for wild-type, fin and fab plants were performed as 52 erence alleles to S. pimpinellifolium and self-fertilizing the F1 plants. fas was described . Briefly, total RNA was extracted from 30–50 meristems per backcrossed to M82 six times and to S. pimpinellifolium ten times, using PCR biological replicate using a PicoPure RNA Extraction kit (Arcturus). mRNA was genotyping markers that identify the left and right breakpoints of the inver- purified from 1–5 µg of total RNA for library construction using the ScriptSeq sion39. lc was also backcrossed to S. pimpinellifolium ten times. All double v2 RNA library preparation kit (Epicentre) with barcodes. The quantity and mutants were generated using standard crossing schemes and were confirmed size distribution of each individual RNA sequencing library were detected on a by genotyping. Seeds used for meristem measurements, RNA sequencing and Bioanalyzer 2100 (Agilent Technologies). Two biological replicates per genotype RNA in situ hybridization were pregerminated in petri dishes on Whatman were generated. Five barcoded libraries were pooled with relatively equal paper moistened with sterile water. Seedlings with similar germination staging concentrations (determined by qRT-PCR quantification) for one lane of Illumina were transferred to soil in 72-cell plastic flats and grown in the greenhouse. paired-end 100-bp sequencing on an Illumina HiSeq sequencing machine. See Seeds used for transient expression were directly sown and germinated in soil methods of data analysis in the Supplementary Note. in 72-cell plastic flats and grown in the greenhouse. Greenhouse plants were Tissue-specific expression patterns for SlCLV3, SlCLE3, SlCLE9, FAB, FIN, grown under natural light with supplementation from high-pressure sodium FAB2 and SlRRA3a were obtained from the tomato tissue RNA sequencing bulbs (50 mM/m2/s) on a 16-h light/8-h dark photoperiod. Daytime and night- database. RNA sequencing data from different tissues (for example, in Fig. 2f,g) time temperatures were 26–28 °C and 18–20 °C, respectively. were mined from the tomato genome project transcriptome profiling data sets deposited in the Sequence Read Archive (SRA) under accession SRP010775 CRIPSR/Cas9 gene editing and genotyping, and phenotyping of the resulting (ref. 53) and from our meristem maturation atlas52. mutants. CRISPR/Cas9 mutagenesis, plant regeneration and greenhouse care were performed as described29. Briefly, constructs were designed to produce defined In situ hybridization. RNA in situ hybridization was performed using deletions within each target gene-coding sequence using two sgRNAs alongside standard protocols54, with slight modifications. Briefly, full-length coding the Cas9 endonuclease gene (see Supplementary Table 6 for a list of the sgRNAs sequences for tomato SlWUS (Solyc02g083950), SlCLV3 (Solyc11g071380) used in this study). For genotyping of each first-generation (T0) transgenic line, and FAB (Solyc04g081590) were amplified from M82 cDNA using Phusion three different leaf samples were collected to capture all possible induced mutant Taq (Invitrogen), and the resulting products were ligated into the StrataClone alleles due to sectoring (chimerism), and genomic DNA was extracted using a pSC-A-amp/kan vector (Agilent Technologies). Plasmids were linearized and, standard cetrimonium bromide (CTAB) protocol. Each plant was genotyped depending on the orientation of the gene insert, T7 or T3 RNA polymerase was

Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature by PCR for the presence of the Cas9-sgRNA1-sgRNA2 construct with primers used for in vitro transcription. Full-length probes were used for hybridization. designed to amplify a region spanning the 3′ end of the 35S promoter and the SAM stages were defined as described previously52. For fixation, meristems 5′ end of Cas9. The CRISPR/Cas9 T-DNA–positive lines were further genotyped for were dissected by hand and fixed in 4% paraformaldehyde with 0.3% Triton indel mutations using a forward primer to the left of sgRNA1 and a reverse primer X-100 under vacuum for 20 min at 400 mm Hg. npg to the right of sgRNA2 (Supplementary Table 6). PCR products from selected plants were purified for cloning into the pSC-A-amp/kan vector (Stratagene). Map-based cloning. Markers for recombinant mapping and genotyping A minimum of eight clones per PCR product were sequenced. Homozygosity were generated using SNPs and indels identified in comparison of 53 for the desired deletion occurred but was rare. We typically obtained biallelic or Solanum lycopersicum and S. pimpinellifolium . By using 76 F2 mutants and chimeric plants in which indels occurred at one or both sgRNA target sites. Only molecular markers derived from the S. pimpinellifolium genome sequence, we plants in which all sequenced alleles were mutated were phenotyped, to ensure mapped the locus for fab-e0497 to a 325-kb interval containing 45 genes on that quantification and comparison of fasciation were based on effectively null the bottom of chromosome 4, including Solyc04g081590, the closest homolog mutants. For quantification of the floral organs of the genotyped mutants, we of CLV1. Using a similar approach with 624 F2 mutants, we mapped the locus randomly collected about 4 flowers from each mutant and counted floral organs; for the fin-e4489 mutant to a 1-Mb region in the middle of chromosome we then pooled the data from 3–5 different mutants to compare with the floral 11 containing 71 genes. As a lack of recombination prevented further fine organs of wild-type plants grown under the same condition. mapping, meristem-enriched mRNA was extracted from fin-e4489 for deep cDNA sequencing (mRNA-seq). A total of 42 million paired-end 100-bp Meristem imaging. Stereomicroscope imaging of meristems was performed Illumina sequencing reads were aligned to the 1-Mb interval, and a single as described previously52. Briefly, hand-dissected tomato meristems were nonsense mutation was detected in Solyc11g064850. Sanger sequencing of captured on a Nikon SMZ1500 microscope. For SEM, meristems were dis- fin-e4632 identified a missense mutation in the same gene. The other fin alleles sected by hand and passed through a dehydrating ethanol series, dried using (fin-n2326, fin-e9501 and fin-n5644) could not be amplified by PCR, indicating a critical-point dryer (Tousimis), coated in gold particles and imaged under a that there were large deletions encompassing the Solyc11g064850 gene location, Hitachi S-3500N microscope. which was confirmed by RT-PCR. Using 42 F2 mutants, we mapped the locus for the fab2-e4981 mutant to a 100-kb region on the bottom of chromosome 8 Transmission electron microscopy. Tomato vegetative meristems from containing 7 genes. Sanger sequencing identified a missense mutation in 9-d-old seedlings were fixed in 2% glutaraldehyde and 2% paraformaldehyde Solyc08g075340. Sequencing of the second allele of fab2, fab2-e0051, identi- in 0.1 M PBS overnight and then post-fixed in 1% osmium tetroxide in 1.5% fied a nonsense mutation in Solyc08g075340.

doi:10.1038/ng.3309 Nature Genetics Transient expression in tomato protoplasts. To generate the transient formate buffer (pH 4.5) overnight at 37 °C, heat inactivated for 10 min at expression constructs for analysis of subcellular localization, eGFP was fused 99 °C and centrifuged for 10 min at 20,800g; the supernatant was collected as to the C termini of FIN, FAB2 and SlRRA3a. To produce a series of Golgi a hemicellulosic/cellulosic extract. The remaining pellet was treated with 1 ml markers, mCherry was fused to the C termini of Man49 (Arabidopsis α-1,2 of saturated Ba(OH)2 (~0.22 M) solution for 6 h at 105 °C. The pellet resulting mannosidase I-49)55, MUR3 (Arabidopsis β-1,2-galactosyltransferase)56 and from centrifugation was washed three times with 1 ml of water per wash and ST (Rattus norvegicus α-2,6-sialyltransferase)57 to create cis-, medial- and centrifuged. The resulting supernatants (extensin fraction) were pooled and trans-Golgi markers, respectively. Fusion protein expression was driven by neutralized with 1 M sulfuric acid, and the precipitating salts were removed the cauliflower mosaic virus (CaMV) 35S promoter in transient expression by centrifugation. The remaining pellet (residue) and all fractions were freeze vectors, as described previously58. The transient expression protocol was dried and analyzed for monosaccharide composition as described above. modified from the Arabidopsis mesophyll protoplast system59. Well-expanded leaves from 10- to 14-d-old tomato plants were cut into 0.5- to 1-mm leaf In planta peptide treatment of tomato apices. Peptide treatment of tomato strips and digested with enzyme solution (1.5% cellulase R10 (Yakult Honsha), apical meristems was modified from previous methods26,62. Sterilized tomato 0.4% macerozyme (Yakult Honsha), 20 mM MES (pH 5.7), 0.4 M mannitol, seeds were distributed onto two layers of Whatman paper moistened with 20 mM KCl, 10 mM CaCl2 and 0.1% BSA) in a 90-mm Petri dish for 4 h in the sterile water in a 180-mm petri dish. The seeds were pregerminated at 28 °C dark with gentle shaking at 40–50 rpm. The digestion products were filtered for 1–2 d. Seeds with similar staging were selected for further growth in a using 75-µm nylon mesh (Fisher) in a 50-ml Falcon tube. Cells were collected 50-ml Falcon tube (10 seeds/tube) with 10 ml of liquid medium containing by centrifugation at 100g for 2 min. The cells were washed by resuspending half-strength Murashige and Skoog salt mixture (Sigma-Aldrich), 1% sucrose them in precooled W5 solution (2 mM MES, pH 5.7, 125 mM CaCl2, 154 mM and 0.5 g/l MES (pH 5.8) and supplemented with different concentrations of NaCl, 0.1 M glucose and 5 mM KCl), and cells were finally resuspended in peptides. The Falcon tubes were incubated on a roller bank with a 16-h light/8-h MMG solution (4 mM MES, pH 5.7, 0.5 M mannitol and 15 mM MgCl2) to a dark photoperiod at 28 °C for 10 d. Chemical synthesis of arabinosylated SlCLE concentration at 2 × 105 cells/ml for immediate transfection. For transfection, peptides was performed as described previously26. 5–10 µg of plasmid of 5–10 kb in size was added to 100 µl of protoplast cell suspension in a 2-ml microfuge tube, and samples were then gently mixed Tomato shoot apical meristem imaging and size measurement of treated well with an equal volume of PEG solution (40% PEG in 0.2 M mannitol meristems. After peptide treatment, SAMs were cleared for imaging and size and 100 mM CaCl2). The transfection mixtures were incubated at room quantification. Individual SAMs from each treatment were dissected and fixed temperature for 10 min, and cells were then washed with W5 solution. in FAA (1% formaldehyde, 0.5% acetic acid and 50% ethanol) overnight at Cells were resuspended in W5 solution and incubated at 28 °C for 12 h before 4°C. The fixed SAMs were sequentially washed with 70%, 85% and 100% collection for microscopy or other experiments. Imaging was performed on ethanol for 30 min each wash and dehydrated for 60 min in 100% ethanol; a Zeiss Axioplan 2 fluorescence microscope with an AxioCam camera. they were then immersed in an ethanol:methyl salicylate solution (1:1) for an additional 60 min. Fixed tissue was transferred to methyl salycilate (100%) for Monosaccharide analysis of the cell wall. Preparation and analysis of cell wall storage before microscopy. Slides were sealed with Vaseline to keep the SAMs material (alcohol-insoluble residue, or AIR) was performed as described60. immersed in methyl salicylate. Imaging by Normarsky illumination was con- Neutral sugars and uronic acids from non-cellulosic polysaccharides in AIR ducted using a Leica DMRB microscope (Leica Microsystems). Micrographs were released by hydrolysis with trifluoroacetic acid (TFA). Approximately were acquired with a Leica MicroPublisher 5.0 RTV digital camera system 1 mg of AIR was hydrolyzed with 250 µl of 2 M TFA at 121 °C for 90 min. The using Q-IMAGING software. All images were taken at the focal plane TFA-soluble material was split into two equal parts, one for uronic acid deter- corresponding to the median optical section of the SAM. The size of each mination and one for neutral sugar determination. Uronic acid content was SAM was determined by outlining and calculating the meristem area between determined by subjecting the hydrolyzate to a CarboPac PA200 anion-exchange the basal edges of two opposite leaf primordia using ‘freehand selections’ and column with an ICS-3000 Dionex chromatography system. The elution ‘measure’, respectively, in ImageJ software. The peptide treatment assays were

Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature profile consisted of a linear gradient of 50–200 mM sodium acetate in 0.1 M performed three times with similar results. NaOH in 10 min at 0.4 ml/min. Neutral sugar composition was determined by derivatization with alditol acetate as described61. In brief, monosaccha- Quantitative RT-PCR. qRT-PCR was performed as described previously52. rides in the hydrolyzate were reduced with NaBH4 (20 mg/ml dissolved Briefly, total RNA from vegetative meristems and shoot apices was extracted npg in 1 M NH4OH) at room temperature for 1 h and then peracetylated with with the PicoPure RNA Extraction kit and the RNeasy mini kit (Qiagen), acetic anhydride and pyridine at 121 °C for 20 min. The alditol acetates gener- respectively. One microgram of total RNA was treated with DNase I and ated were separated on a gas chromatograph equipped with a Supelco SP2380 used for cDNA synthesis with a SuperScript III reverse-transcriptase kit column (Sigma-Aldrich) and analyzed by a flame ionization detector (FID) (Invitrogen). qRT-PCR was performed using gene-specific primers in the iQ (Agilent Technologies, 7890A). SYBR Green SuperMix (Bio-Rad) reaction system on the CFX96 Real-Time system (Bio-Rad), following the manufacturer’s instructions (Supplementary Sequential extraction of cell wall polymers. Cell wall polymers were sequen- Table 7). The tomato UBIQUITIN gene was used as an internal control. tially extracted as described previously31. Freeze-dried plant materials were ground to fine powder. About 5 mg of ground material was extracted overnight Constructs for transformation and complementation of Sp-fas. The binary in 50 mM ammonium formate buffer, pH 4.5, at 37 °C with shaking at 200 rpm. vector (pHaoNM) used in transgenic complementation of Sp-fas was modified After incubation, the samples were centrifuged for 10 min at 20,800g, and from the standard vector pCAMBIA1300. The two constructs used for com- the supernatant was collected. The remaining pellet was washed three times plementation tests were generated through the following steps. First, fosmid with 1 ml of water per wash and centrifuged, and the resulting supernatants SL_FOS0151E08 from tomato accession Heinz 1706 was digested at MfeI were pooled with the supernatant from the first spin. The supernatant was and SacI sites to obtain a 4,320-bp fragment containing the SlCLV3 gene. precipitated overnight in 80% ethanol and centrifuged for 10 min at 20,800g. This fragment was then ligated into pHaoNM at the EcoRI and SacI sites The precipitated fraction was freeze dried, representing the buffer extract. (pHC1). A 5,233-bp fragment upstream of the SlCLV3 gene was isolated from The buffer-insoluble pellet was digested with 0.02 U pectin methylesterase SL_FOS0151E08 by digesting at SacI and NsiI sites. This fragment was ligated (EC3.1.1.11, Novozymes) and endopolygalacturonanase M2 (EC3.2.1.15, into pHC1 at the SacI and SbfI sites to create the complementation construct Megazyme) in 1 ml of 50 mM ammonium formate buffer (pH 4.5) overnight named long-gCLV3, containing the SlCLV3 gene with 5,465 bp upstream of the at 37 °C. After digestion, the samples were treated for 10 min at 99 °C and start codon and 3,445 bp downstream of the stop codon. For the control con- centrifuged for 10 min at 20,800g, and the supernatant was collected as the struct short-gCLV3, a PCR product was amplified with primers 12EP315 and EPG/PME fraction (pectin extract). The remaining pellet was digested with 12EP389 (Supplementary Table 7). The PCR product was then digested with 10 U endoglucanase II (EGII, Megazyme) in 1 ml of 50 mM ammonium SacI and PstI and inserted into pHC1at the SacI and SbfI sites. The resulting

Nature Genetics doi:10.1038/ng.3309 short-gCLV3 construct contained the SlCLV3 gene, 1,008 bp upstream of the 53. Tomato Genome Consortium. The tomato genome sequence provides insights into start codon of SlCLV3 (up to the fas inversion breakpoint) and 3,445 bp down- fleshy fruit evolution. Nature 485, 635–641 (2012). 54. Jackson, D.P. in Molecular Plant Pathology: A Practical Approach (eds. Gurr, S.J., stream of the stop codon. Bowles, D.J. & McPherson, M.J.) 163–174 (Oxford University Press, 1992). To test whether SlCLV3 was responsible for the fas phenotype, long-gCLV3 55. Saint-Jore-Dupas, C. et al. Plant N-glycan processing enzymes employ different and short-gCLV3 were introduced separately into Sp-fas using Agrobacterium- targeting mechanisms for their spatial arrangement along the secretory pathway. mediated transformation (Agrobacterium tumefaciens strain LBA4404). Plant Cell 18, 3182–3200 (2006). 56. Chevalier, L. et al. Subcompartment localization of the side chain xyloglucan- We generated 18 first-generation (T0) transgenic plants, which all carried synthesizing enzymes within Golgi stacks of tobacco suspension-cultured cells. Plant 1 or 2 insertions of long-gCLV3. All 18 plants were used for phenotypic analysis J. 64, 977–989 (2010). of fruit locule number (n 200). We also generated 18 T0 plants carrying short- 57. Nelson, B.K., Cai, X. & Nebenfuhr, A. A multicolored set of in vivo organelle markers gCLV3, and we phenotyped 16 plants that carried 1 or 2 insertions. Two plants for co-localization studies in Arabidopsis and other plants. Plant J. 51, 1126–1136 (2007). that were found by Southern blot to carry four and six copies of the transgene 58. Xu, C. et al. Degradation of MONOCULM 1 by APC/CTAD1 regulates rice tillering. showed evidence of cosuppression and were therefore excluded from analysis. Nat. Commun. 3, 750 (2012). A more detailed analysis of complementation of fas with long-gCLV3 was per- 59. Yoo, S.D., Cho, Y.H. & Sheen, J. Arabidopsis mesophyll protoplasts: a versatile cell system for transient gene expression analysis. Nat. Protoc. 2, 1565–1572 formed in T1 progeny plants derived from a T0 plant with a single transgene (2007). insertion. Rescue of bilocular fruit cosegregated with the presence of the long- 60. Xiong, G., Cheng, K. & Pauly, M. Xylan O-acetylation impacts xylem development gCLV3 transgene, whereas short-gCLV3 had no effect (Fig. 6). Genotyping for and enzymatic recalcitrance as indicated by the Arabidopsis mutant tbl29. Mol. the presence of the transgene in T1 progeny plants was performed using PCR Plant 6, 1373–1375 (2013). primers that amplified a fragment of the pCAMBIA1300 backbone. 61. York, W.S., Darvill, A.G., McNeil, M. & Albersheim, P. 3-deoxy-d-manno-2-octulosonic acid (KDO) is a component of rhamnogalacturonan II, a pectic polysaccharide in the primary cell walls of plants. Carbohydr. Res. 138, 109–126 (1985). 52. Park, S.J., Jiang, K., Schatz, M.C. & Lippman, Z.B. Rate of meristem maturation 62. Fiers, M. et al. The CLAVATA3/ESR motif of CLAVATA3 is functionally independent determines inflorescence architecture in tomato. Proc. Natl. Acad. Sci. USA 109, from the nonconserved flanking sequences. Plant Physiol. 141, 1284–1292 639–644 (2012). (2006). Nature America, Inc. All rights reserved. America, Inc. © 201 5 Nature npg

doi:10.1038/ng.3309 Nature Genetics